Skip to content

release: Memmy v1.1.2 - #338

Draft
syzsunshine219 wants to merge 87 commits into
mainfrom
release/v1.1.2
Draft

release: Memmy v1.1.2#338
syzsunshine219 wants to merge 87 commits into
mainfrom
release/v1.1.2

Conversation

@syzsunshine219

@syzsunshine219 syzsunshine219 commented Sep 1, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Prepare Memmy v1.1.2 while preserving the independently versioned Memory runtime and CLI at 2.1.0.
  • Integrate the current official release/v1.1.2 tip and current main with a normal merge. The merge result preserves the release tree while recording main as the second parent.
  • Resolve the three onboarding/BYOK test conflicts in favor of the release branch's installation-scoped first-report semantics.
  • Fix full Agent-source rescans so persisted messageCount reflects the complete scan instead of accumulating duplicate counts; align the executable smoke with the version-2 executor state that no longer stores legacy request-ID arrays.

Final source

  • Final HEAD: c7feae01a109cb1076b305ef1a364d348b808e02
  • Main integrated: e679cf9caff48c59f03bb5587d35152cde013988
  • Previous release tip: f9b1b95d7833ad8c5e9de024e658a20ce7b6c0b6
  • Merge commit: eee946003a2c14f746402a66af3c8589f32177a2
    • parents: f9b1b95d7833ad8c5e9de024e658a20ce7b6c0b6 and e679cf9caff48c59f03bb5587d35152cde013988
    • merge tree equals the previous release-tip tree exactly
  • Follow-up regression fix: c7feae01a109cb1076b305ef1a364d348b808e02

Verification

All local commands used Node v22.23.2 (ABI 127). Both root and independent App/memmy-agent npm ci completed successfully from their committed lockfiles. No selected check was waived.

  • Conflict regressions: Backend 53/53; Frontend 117/117.
  • Release workflow: 27/27. Packaging guards: 26/26.
  • Linux local contracts: archive/installer 5/5, Agent gateway/systemd 41/41, Memory CLI setup 22/22.
  • Memory: strict typecheck; 90 files / 714 tests; source-integration additions 17/17.
  • Backend: strict typecheck; 109 files / 834 tests.
  • Frontend: strict typecheck; 163 files / 1,539 tests.
  • Agent: strict typecheck; 294 files passed / 1 skipped, 5,126 tests passed / 3 skipped.
  • Desktop shell: strict typecheck; 21 files passed / 5 skipped, 221 tests passed / 106 skipped.
  • AgentSourceCore: 4/4. Migrations: 117 passed / 1 skipped.
  • Local API and storage/migration focused matrices passed: 22/22, 77/77, 68/68, 23/23, and 7/7.
  • Root lint, typecheck, npm test, root build, independent Agent build, and final version:check passed.
  • Actual localhost/temp-directory smokes passed for Memory/Backend, Codex + Cursor + Claude ingestion, and Cursor-only routing.

The actual smoke first reproduced two stale/full-rescan failures. Regression coverage was added before the minimal fixes:

  1. The smoke expected removed importedRequestIds arrays instead of version-2 source checkpoints.
  2. Repeated mode=full scans doubled persisted source message counts from 2 to 4 despite Memory deduplication.

The final implementation requires executor state version 2, rejects persistence of legacy ID arrays, compares stable checkpoint fields, and only accumulates message counts for incremental scans.

Project Harness evidence

  • Task: feature-2026-09-01-ff95d3c3-243d0c35
  • Revision: 1
  • Route: T3 / Full
  • Evidence run: run-1788423267330-af45bb03-53a2-467b-9f98-e9000b101d44
  • Evidence fingerprint: 86c44c6ae10b51930b38cf56fb4d7f0157d6e1b11ffa2eeb01abf3c854e90b10
  • Selection fingerprint: 5ce344d54761c97a86baa8729034553ebc0c9a1d4de9726da3bc913c350d82c1
  • Manifest fingerprint: af2792fe3f3a1ccd8bab6051002e38c05b657b7d1d46ed4b6deb330bafef08b7
  • Full evidence: 12/12 passed
  • Gate: needs_review; verificationPassed=true; deliveryReady=true; failedEvidence=[]

Linux provider

  • PR run 33732597378 is bound to final HEAD c7feae01a109cb1076b305ef1a364d348b808e02 and succeeded:
    • build: success
    • native x64 install-twice/CLI smoke: success
    • native arm64 install-twice/CLI smoke: success
    • release upload: skipped
  • Explicit non-release dispatch 33732611085 also succeeded with upload_to_release=false.

Existing Desktop artifact provenance

Four signed/notarized macOS DMGs were inspected. Their embedded Desktop version (1.1.2), Memory runtime version (2.1.0), edition, account channel, native architecture, Developer ID signature, Gatekeeper status, and stapled notarization ticket are internally consistent.

However, none has a neighboring release manifest, checksum/build log, embedded source commit, or other record binding it to final HEAD c7feae01a109cb1076b305ef1a364d348b808e02. They are therefore provenance-stale and must not be used as final revision-1 release artifacts. The trusted packager must rebuild all required Desktop artifacts from the final HEAD and retain the source-SHA-bound manifest and artifact SHA-256.

Boundaries

  • This PR remains Draft and requires independent T3 review by a different eligible MemTensor maintainer after the new content-bound Harness handoff marker.
  • This work does not authorize PR merge, approval, tag creation, GitHub Release creation, publishing, Desktop package construction, signing, notarization, release upload, OSS upload, or real-device acceptance.

hijzy and others added 30 commits August 25, 2026 16:32
feat(cli): make onboard interactive by default
feat(memory): add standalone Memory 2.1 runtime, multi-agent ingestion, and Viewer
fix(onboarding): prevent first report output leaks
….1.2

feat(settings): refine token usage and model configuration
fix: harden model compatibility and agent indexing
@syzsunshine219

Copy link
Copy Markdown
Collaborator Author

Project Harness review handoff marker. Only reviews submitted strictly after this GitHub timestamp qualify.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants